We have determined structures of IP kinases from two archaeal species, THA and MTH, with crystals obtained under a variety of conditions. All crystals were grown in the presence of IP and ATP. The THA IPK structure was determined by single wavelength anomalous diffraction (SAD) using crystals of selenomethionine-substituted protein grown in 0.1 M citrate buffer, pH 5.0. The asymmetric unit contained two molecules of IPK, each of which contained IP and ADP in the active site. This combination of ligands presumably resulted from hydrolysis of ATP that occurred during sample preparation and crystallization. This refined structure was used as a search model in subsequent molecular replacement solutions of other IPK structures. THA IPK crystals grown in 0.1 M malonate/imidazole/borate (MIB) buffer, pH 7.0, also contained two molecules of the enzyme in the asymmetric unit but each contained the substrates IP and ATP in the active site (IPK•IP•ATP). In addition, THA IPK crystals grown in propionate/cacodylate/Bis-Tris (PCB) buffer, pH 6.0 contained four molecules in the asymmetric unit where one molecule contained IP and ATP while the other three contained the products, IPP and ADP, formed during crystallization. The presence of substrate and product complexes in the asymmetric unit is reasonable given an equilibrium constant for the reaction, Keq
= 6.3, for uncomplexed substrates and products (13
). Finally, a structure for MTH IPK in complex with glycerol and ADP (IPK•GOL•ADP) was determined. All of these structures were refined to Rfree
values of 23–26% against data to 2.0–2.15 Å resolution. All four structures have excellent geometry, with more than 95% of the residues in favored Ramachandran regions. Crystallographic statistics are given in .
All of the crystallized proteins of THA IPK included N-terminal Phe(−1) and Thr(0) residues that were the remnants of a TEV protease recognition site. In some cases, one or both of these residues were visible in the electron density. Moreover, electron density for residues Asp(−3) and Pro(−2), also parts of the TEV protease recognition site were found in the asymmetric unit of the crystal grown in PCB buffer. This asymmetric unit was interesting because of the combination of substrates and products in the molecules mentioned earlier, as well as the significant movement of loops in two molecules. Briefly, molecules A, B and C contained bound products, while molecule D contained bound substrates. Molecules B and D formed the only homodimer in the asymmetric unit, resulting in an interesting mix of ternary enzyme complexes in one dimeric structure. In molecules C and D, the η1-αA loop was in a conformation different than those in molecules A and B, which allowed these molecules to make crystal lattice contacts through Tyr17 (Supplementary Figure 1
, panel a). This loop movement resulted in a significant shift in the side chain position of Lys14, an invariant residue in IPK’s with direct involvement in catalysis that will be discussed later. Moreover, in molecules C and D, alternate conformations were found for another catalytic residue, His50 (Supplementary Figure 1
, panel b). An active site water molecule found in enzymes structurally related to IPK was found only in molecules A, C and D. The unique characteristics of these asymmetric unit molecules allowed us to choose molecule A as the best structure that represents the THA IPK product complex (IPK•IPP•ADP).
Fold and quaternary structure of IPK
IPK belongs to the amino acid kinase (AAK) family of enzymes that also includes N-acetyl-L-glutamate kinase (NAGK) (14
), the aspartokinase (AK) N-terminal domain (15
), glutamate-5-kinase (G5K) (16
), carbamate kinase (CK)(17
), uridine monophosphate kinase (UMPK) (18
) and fosfomycin kinase (FomA) (19
). This family is characterized by the open αβα sandwich fold first observed in E. faecalis
) and catalyzes the transfer of a phosphoryl group from a nucleotide to a small molecule acceptor. Depending on the functional group that is phosphorylated, this family can be divided into two subdivisions. Enzymes in the “carboxylate” subdivision catalyze the transfer of a phosphoryl group to a carboxylate or carbamate group (NAGK, CK, G5K and AK), while those in the “phosphate” subdivision transfer a phosphoryl group to a phosphate or phosphonate (UMPK, FomA, IPK). THA IPK is structurally most similar to FomA, where the backbone RMSD is 2.6 Å on 224 residues. Overlap of FomA and NAGK with THA IPK also aligns their respective substrates and most of their active site residues, suggesting similar catalytic mechanisms. As expected, MTH IPK shares high structural homology with THA IPK with RMSD of 1.9 Å over 223 Cα atoms, although sequence identity is only 32%. These structural comparisons confirm the correct assignment of THA and MTH IPK’s to the AAK family of enzymes.
The THA IPK structure contains an N-terminal domain (residues 1–155) that binds IP and a C-terminal domain (residues 156–245) that binds ATP. Each molecule of THA IPK consists of 16 β-strands, eight α-helices and two 310
helices interconnected by loops and turns (, panel a). The open αβα sandwich fold formed by these secondary structures is highly similar to that of NAGK (14
): the same structural elements form the central β-sheet core (β5, β8, β2, β1, β11, β15, β16, β14) and the two flanking layers of α-helices (αC, αA, αH on one side and αD, αE, αG, αF on the other). Three β-hairpin turns reside in the N-terminal domain, two of which form the surrounding walls of the aliphatic tail of IP (β3–β4, β9–β10) and another forming one side of the IP binding pocket (β6–β7). The β3–β4 hairpin turn of THA IPK is shorter than that of NAGK but is in the same closed conformation that allows it to cap the IP binding site and shield its hydrophobic surroundings from solvent molecules (14
). The remaining β12–β13 hairpin turn resides in the C-terminal domain and hovers above the purine ring of ATP. This β-hairpin is found within the segment that connects the core β-sheet strands β11 and β14 and is longer than the corresponding segment in NAGK. The longest helix in THA IPK, αC, aligns well with αC in NAGK, but is one turn longer.
Figure 2 Tertiary structure of THA IPK. (a) A ribbon diagram of the THA IPK monomer with the helices and strands labeled. The region that connects β14 and αG is disordered in the crystal and not modeled in the structure. (b) The THA IPK primary (more ...)
The segment between β14 and αG (residues 189–201) in the THA IPK structure is disordered in all molecules in the asymmetric unit, as evidenced by an un-interpretable electron density. This segment contains the helix, αF and the αF–αG junction loop that form one side of the ATP binding site. The corresponding segment in MTH IPK is missing (204–218), as is also seen in one molecule of G5K (2J5T) (16
), suggesting that the flexibility of these secondary structures might be important for the binding and release of ATP.
The THA and MTH IPK’s are dimeric in the crystal, consistent with the observed oligomerization state during gel filtration chromatography (, panels and b).The dyadic axis is perpendicular to the core β-sheet that runs across the dimer (16 β-sheets, 8 per subunit). As in all enzymes of this family that catalyze phosphoryl transfer, each active center is confined to a single subunit. However, dimer formation may be structurally significant since N-terminal domain loops surrounding the active site are anchored to the other subunit of the dimer (14
Figure 3 Homodimers of THA and MTH IPK’s. (a) The THA IPK homodimer, with secondary structures involved in the dimer interface labeled. (b) The MTH IPK homodimer. The dyadic axis is perpendicular to the core β-sheet that runs across the dimer and (more ...)
Analysis of the dimerization surface of THA IPK using the EBI PISA (20
) server resulted in a complexation significance score of 0.797, corresponding to a tight dimer that buries a total of 3275 Å2
of solvent accessible area between the two monomers. The compactness arises from a total of nine hydrogen bonding interactions between residues in the long interfacial αC helices of both subunits and between the αD helix of one subunit and the β4, β6, and β9 strands and the η2 helix of the other. Eight electrostatic interactions between the αC helices of the two subunits and between the β6 strand of one subunit and the β7 strand of the other further stabilize the dimer. Finally, hydrophobic interactions mediated by residues in the αC, αD and η3 helices, and the β4, β5, β6, β9 and β10 strands further bind the two molecules.
Although essentially the same secondary structures are used by the AAK enzymes in dimerization, different quaternary architectures are generated by changing the angle between the long interfacial helices αC of the dimer subunits (16
). In the case of THA IPK, the crossover point of these helices is found on the fifth turn from the N-terminus, and the rotation angle around an axis that penetrates the inter-subunit interface perpendicularly at this point is ~220°. In MTH IPK, the crossover point shifts to the sixth turn from the N-terminus while the rotation angle remains at ~220°. In NAGK and CK the crossover points in αC occur in the third turn, and shifts to the fifth turn in G5K, while the rotation angles are 110° in both NAGK and CK and 260° in G5K (16
The active site
The THA IPK structures presented here are the first to show the molecular details of binding IP and IPP molecules by an enzyme (, panels a and b). Many characteristics of AAK enzymes have been described previously in the discussion of the NAGK structure, and several of these traits are conserved in both THA and MTH IPK’s. Briefly, the IP (IPP) binding pocket is located in the N–terminal domain formed by the β2–αB loop, the αB helix, the β3–β4 hairpin, the N–terminus of the αC helix, the β8–β9 loop, the β9–β10 hairpin turn, and the β10–αE loop. The αB helix contributes to binding IP (IPP) through the orientation of the positive end of the helix dipole and the hydrogen bonding and electrostatic interactions made by the charged His50 residue. In the FomA structure, the αB helix is ordered in the presence of fosfomycin and disordered in its absence (19
). Comparison of AAK enzyme structures revealed that only the enzymes in the “phosphate” subdivision (UMPK, FomA and IPK) align their respective αB helices. Moreover, an aligning residue at the His50 position is found exclusively in these enzymes.
Figure 4 IP and IPP binding sites (a) The IP binding site showing the amino acid residues that form the hydrophobic interior of the pocket and limit the chain length specificity of the enzyme. Hydrogen bonds between active site residues and the phosphate moiety (more ...)
The His50 residue in IPK•IP•ATP poise the bound IP for nucleophilic attack on Pγ
of ATP by hydrogen bonding with a non-bridging oxygen atom in the terminal phosphate of IP. In the IPK•IPP•ADP structure, this residue moves slightly to make a hydrogen bond with a non-bridging oxygen atom in the IPP terminal phosphate (, panel b), facilitating the reverse reaction. In fact, in comparing IPK•IP•ATP and IPK•IPP•ADP (r.m.s.d. 0.4 Å), the slight change in the position of His50 to hydrogen bond with the terminal phosphate of the product is one of the very few coordinate shifts that were observed. In FomA, His58 positions fosfomycin for nucleophilic attack on the phosphate donor through hydrogen bonding with an intervening water molecule (19
). The corresponding residue in E. coli
UMPK, Arg62, hydrogen bonds with the terminal phosphates of both substrate and product and stabilize Pγ
of ATP through electrostatic interactions (21
). These observations together with the invariance of this histidine residue in putative IPK homologues (13
, Supporting Information, Table S2
), the apparent pKa of 6.5 for THA IPK (6.8 for MTH IPK) from pH dependence studies (13
) suggest that this His50 may perform direct catalytic roles similar to those proposed for Arg62 in UMPK, besides being a signature trait of IPK’s.
The specificity of IPK arises from hydrophobic residues that cradle the aliphatic portion of IP and a constellation of conserved residues that hydrogen bond with its phosphate moiety (, panel a). The residues Met77 and Val73 (αC N-terminus), Val130 and Ile140 (β9–β10 hairpin) and Ala53 (αB C-terminus) and Tyr127 (β8–β9 loop) all form the hydrophobic pocket for the apolar tail of IP. In addition, Leu58 and Pro59 of the β3–β4 hairpin enhance the hydrophobic surroundings by capping the IP binding pocket. Notably, the β3–β4 hairpin is found only in the structures of NAGK and IPK. This motif is in the closed conformation in substrate-bound NAGK structures (14
) as in the IPK•IP•ATP and IPK•IPP•ADP structures. The side chains of some, if not all, of these hydrophobic residues may give rise to the chain length specificity of IPK. Geranyl phosphate, which is one isoprene unit longer than IP, is a poor substrate for IPK, indicating that the IP binding site can not easily accommodate a C10
isoprenoid chain (13
The phosphate moiety of IP is recognized in the active site by structural motifs that are well-conserved in the AAK family. In IPK•IP•ATP, the three non-bridging oxygen atoms of IP make hydrogen bonding interactions with the Nε
of His50, the N atom of Gly45 in the β2-αB loop, and Asp144 through a bound water molecule (, panel a). This water molecule is found in the holoenzyme structures of E. coli NAGK (pdb ID 1GS5), P. furiosis UMPK (pdb ID 2BMU), E. coli UMPK (pdb ID 2BNE) and FomA (pdb ID 3D41). A water molecule in a similar position is present in IPK•IPP•ADP (, panel b). The β2-αB loop, which contains Gly45, is one of the conserved glycine-rich loops in the AAK family and is thought to stabilize the transition state intermediate and the phosphorylated product of these enzymes (19
). In IPK•IPP•ADP, Gly45 of this loop makes a hydrogen bond with the Pα
bridging oxygen atom of IPP. The other glycine-rich loop, β1-η1, contains Gly8 that stabilizes IPP by making a hydrogen bond with a non–bridging oxygen atom of its terminal phosphate.
The β10–αE loop completes the structural elements involved in binding the polar end of IP. This loop contains Ser142, which is conserved in the “phosphate” subdivision (Thr120 in P. furiosis
UMPK, Ser149 in FomA) (18
) and makes a hydrogen bond with the terminal phosphate of the substrates. In IPK•IP•ATP and IPK•IPP•ADP, the orientation of the side chain hydroxyl group of Ser142 does not allow a hydrogen bond with substrates or products (, panel a and b). However, an alternate rotamer could provide hydrogen bond stabilization as the substrates moved to the transition state. The positive end of the helix dipole of αE, initiated by the β10-αE loop, may further stabilize the polar head of the IP molecule as well as the negatively-charged transition state intermediate, as in other structures of AAK enzymes.
Thus, the specificity of IPK towards IP arises from a combination of hydrophobic interactions with the hydrocarbon tail and hydrogen bonding and electrostatic interactions with the phosphate head group to match the amphiphatic character of IP. This differentiates the IPK active site from other enzymes in the family whose substrates do not have this intrinsic amphiphathicity, such as that of NAGK whose active site is laden with both polar and charged residues that interact with the acetyl and carboxyl groups of NAG. This unique binding elements of IPK limits its catalytic promiscuity.
The location of the ATP binding site in both THA and MTH IPK’s superimpose well with those of NAGK and other AAK enzymes. Conserved amino acid residues in these enzymes that have previously been identified to form hydrogen bonding, electrostatic and hydrophobic interactions with the purine ring, sugar and polyphosphate moieties of ATP are also found in IPK. For instance, the NAGK residues Met214 (purine stacking), Asp181 (ribose 3’-OH binding), Asp162 (polyphosphate positioning) and Lys8 (phosphate hydrogen bonding) (14
) are also conserved in THA IPK as Ile202, Asp164, Asp144 and Lys5 respectively. In MTH IPK, the corresponding residues are Met220, Asp178, Asp158 and Lys5. Other ATP-binding residues in THA IPK that superimpose with similar ATP-binding residues in NAGK are shown in ().
Figure 5 Stereo view comparisons of THA IPK and NAGK active sites. THA IPK (green) and E. coli NAGK (red) align 207 Cα with an r.m.s.d. of 3.0 Å. Sequence identity is 14%. The bound molecules of THA IPK (IP and ATP) and NAGK (NAG and AMPPNP) align (more ...)
The β1-η1 glycine-rich loop that contains Gly8 engages the Pβ
of ATP in IPK•IP•ATP through hydrogen bonding. These interactions by the conserved Gly8 with the phosphate groups ATP (and with the terminal phosphate of IPP stated earlier) are seen for the corresponding residue in other enzymes and reflect a role in phosphoryl transfer (21
). The residue Asp172 and Lys174 form hydrogen bonds with the 2’- and 3’-OH groups of ATP. Finally, a hydrogen bond with side chain carboxylate of Asp164 completes an intricate network of interactions that secure this end of ATP, thus positioning the polyphosphate end for nucleophilic attack by IP.
As stated earlier, αF and the loop that connects it to αG were not modeled in our IPK structures. These secondary structures contain Phe191, analogous to Trp202 in FomA, which stacks on the adenine ring, and the conserved DVTGG sequence that make important binding interactions with the nucleotide seen in other AAK enzymes. We suspect that these interactions are more or less conserved in IPK, although a more complete IPK structure would be necessary to unequivocally establish the nature of these interactions. The flexibility and location of these absent secondary structures suggest a role in nucleotide sequestration and release.
The IPK•IP•ATP structure was obtained without resorting to a substitution of the Pβ
bridging oxygen atom that would prevent phosphoryl transfer. In addition, phosphoryl transfer between added substrates during the crystallization experiments provided IPK•IPP•ADP. These two structures, providing snapshots of the IPK enzyme before and after catalysis, suggest an important role for Lys14 in stabilizing the transition state (, panel a and b). This residue is also conserved in FomA and homologs of IPK (13
, Supporting Information, Table S2
). An associative mechanism of phosphoryl transfer was proposed for NAGK (14
). The similarity of the over-all fold and the superposition of the substrate and nucleotide binding sites of THA and MTH IPK’s with those of NAGK suggests a similar mechanism for IPK (). In IPK•IP•ATP, a nucleophilic oxygen atom the phosphate group in IP is 2.9 Å from the electrophilic Pγ
phosphate atom in ATP, and is poised for attack. The products in IPK•IPP•ADP are likewise poised for the reverse reaction, where the nucleophilic oxygen atom in Pβ
of ADP is 2.3 Å from the electrophilic Pβ
phosphate atom of IPP.
Proposed catalytic mechanism for IPK. Catalytic residues and selected hydrogen bonds are shown. The “missing” Mg2+ ion is positioned based on overlap with the structure of FomA.
Figure 7 Electron density maps of the active sites of THA and MTH IPK’s contoured at 1 RMSD. (a) The THA IPK active site with bound substrates, IP and ATP. (b) The THA IPK active site with bound products, IPP and ADP. The Gly8 residue is obscured by P (more ...)
As in NAGK, the conserved glycine rich loops β1-η1 and β2-αB of THA IPK, the positive helix dipoles of αB and αE, the positive charges of Lys5 and Lys205, and Asp144, which positions Lys5 and Lys205 towards the polyphosphates, create an environment for stabilizing the negatively charged transition state (, panels a–c). A divalent metal ion, Mg2+
, further contributes to stabilization by the same mechanism in addition to its effect on the orientation of the ATP polyphosphate (14
). In IPK•IP•ATP and IPK•IPP•ADP, a Mg2+
ion could not be unambiguously assigned to the electron density that superimposes onto the Mg2+
atom in the structure of NAGK. Instead, a water molecule is assigned to this density since the distance of the sphere’s centroid to the relevant polyphosphate oxygen atoms are more consistent with hydrogen bonding than with metal coordination. Nevertheless, product turnover assays using radioactive [32
) and Mg2+
confirmed the divalent metal ion dependence of IPK, a well-established characteristic of kinases (Supplementary Figure 3
, panel a)(23
). IPK was also active in the presence of Mn2+
, and Co2+
(Supplementary Figure 3
, panel b).
In IPK•IP•ATP, Lys14 is located within hydrogen bonding distance of the non-bridging oxygen atoms of Pα
(3.0 Å) and Pγ
(2.8 Å) and the Pβ
bridging oxygen atom (3.1Å) in ATP. In IPK•IPP•ADP, Lys14 makes a hydrogen bond with a non-bridging oxygen atom of Pβ
(2.7 Å) in IPP and the non-bridging oxygen atoms of Pα
(2.7 Å) and Pβ
(3.0 Å) of ADP. Interestingly, a corresponding residue, Lys18, is found in the (FomA•fosfomycin•AMPPNP) complex (19
). However, in this structure, Lys18 only forms a hydrogen bond with a Pγ
oxygen atom in AMPPNP, which in turn, is not properly oriented for nucleophilic attack by fosfomycin. The relative positions of IP and ATP in THA IPK are consistent with phosphoryl transfer. Thus, in THA IPK Lys5, Lys14 and Lys205 form a “lysine triangle” surrounding the phosphate residues in the substrates of IPK’s.
Along with His50, Lys14 is a conserved feature of the IPK’s. These residues may have evolved to preferentially stabilize transition states that contain more negative charge than those formed during phosphoryl transfers involving carboxylated and carbamated substrates, thus giving rise to IP kinases. The conservation of these traits in FomA along with other key catalytic residues in seen in IPK is also consistent with the acquisition of antibiotic resistance toward fosfomycin in Streptomyces by a lateral gene transfer of an ancestral IPK gene from Archaea.